A Learning Automata Based Spectrum Prediction Technique for Cognitive Radio Networks

This paper introduces an application of artificial intelligence in the cognitive radio networks. The Cognitive Radio Network (CRN) provides a suitable environment for Secondary Users (SUs) to share the spectrum with Primary Users (PUs) in a non-interfering manner. In order to determine the availability of PUs bandwidth, SU can sense the spectrum in the channel. But, accurate and constant spectrum sensing consumes the energy of the SUs significantly. In these conditions, to discover the spectrum holes in the absence of PUs, predictive techniques can be one of the solutions which can reduce the consuming energy of the SUs. The simplicity and reliability of predictive techniques play an important role in the practice. In this paper, we utilize a Learning Automata technique to predict the spectrum hole in the cognitive network based on the statistical behaviour of the PUs. Simple structure and acceptable prediction rate are two important features of the proposed technique. In order to compare the performance of the proposed method with similar predictive techniques in CRNs, we design a predictor model using multilayer perceptron artificial neural networks and test the performance of these two methods on the same conditions. The results of modelling confirm that the Learning Automata with simple structure is more reliable than neural network.